[ANNOUNCE] autofs 5.0.1 rc3 release candidate

[ANNOUNCE] autofs 5.0.1 rc3 release candidate

[Ian Kent]

Hi all,

While there are still some issues the number of patches is hard
to manage so I'm consolidating again.

Known issues
============

Quoted strings in the master map are not yet handled.
This may be delayed unlil the first bug fix release following
the initial release.

There will always be bugs so I need to draw a line at some point
and issue a release of 5.0.1. That means this might be the last
oppertunity to test before the release so please find time to
check this out.

autofs
======

The package can be found at:

ftp://ftp.kernel.org/pub/linux/daemons/autofs/v5

It is autofs-5.0.1-rc3.tar.[gz|bz2]

No source rpm is there as it can be produced by using:

rpmbuild -ts autofs-5.0.1-rc3.tar.gz

and the binary rpm by using:

rpmbuild -tb autofs-5.0.1-rc3.tar.gz

See the INSTALL file for information about configure options and
kernel requirements.

Note:
The rc2 tar and patches have been moved to the "old" subdirectory
so any automated build processes that are looking for the tar or
patches in the location above will need to be updated.

Here are the entries from the CHANGELOG which outline the updates:

4/1/2007 autofs-5.0.1 rc3
-------------------------
- fix handling of autofs specific mount options.
- fix include check full patch for file map of same name.
- fix cache entrys not being cleaned up on submount expire.
- fix LDAP lookup delete cache entry only if entry doesn't exist. 
- add missing socket close in replicated host check (Jeff Moyer).
- remove unused option UNDERSCORETODOT from default config files.
- make default installed master map for /net use "-hosts" instead
  of auto.net.
- fix included map recursive map key lookup.
  - and fix the recursive map key lookup for browsable map case.
- review and fix master map options update for map reload.
- fix "-fstype=nfs4" handling.
- fix get_query_dn not looking in subtree for LDAP search.
- allow syntax "--timeout <secs>" for backward compatibility.
- make masked_match independent of hostname for exports comparison.
- fix file handle leak in nsswitch parser.
- fix memory leak in mount and expire request processing.
- add additional check to prevent running of cancelled tasks.
- fix potential file handle leakage in rpc_subs.c for some failure cases.
- fix file handle leak in included map lookup.
- fix "-fstype=nfs4" server probing.
- set close-on-exec flag on open files where possible.
- fix parsing of numeric host names in LDAP map specs.
- fix get_query_dn not looking in subtree for LDAP search (missed second
  occurance).
- allow additional common LDAP attributes in map dn.
- deal with changed semantics of mkdir in 2.6.19.
- fix macro table locking.
- fix nsswitch parser locking.
- allow only one master map read task at a time.
- fix misc memory leaks.
- mitigate manual umount of automounts where possible.
- fix multiply recursive bind mounts.
- check kernel module version and require 5.00 or above.
- fix expire regression introduced in the "mitigate manual umount" patch.
- still more on multiply recursive bind mounts.
- fix tokenizer to distinguish between global option and dn string.
- fix incorrect return from spawn (Gordon Lack).
- fix parsing of bad mount mount point in master map.
- fix use after free memory access in cache.c and lookup_yp.c.
- eliminate use of pthread_kill to detect task completion.
- alter nfs4 host probing to not use portmap lookup and add options
  check for "port=" parameter.
- correct semantics of "-null" map handling.
- remove ability to use multiple indirect mount entries in master map.
- expand export access checks to include missing syntax options.
- make "-hosts" module try to be sensitive to exports list changes.
- change mount "device" from "automount" to the map name.
- check for buffer overflow in mount_afs.c.
- update master map tokenizer to admit "slasify-colons" option (Capelle Bonoit).
- update location validation to accept "_" (Fabio Olive Leite).
- set close-on-exec flag on open sockets.
- fix nonstrict multi-mount handling.
- reduce thread stack to less excessive size.
- update kernel patches.

Ian

Re: [ANNOUNCE] autofs 5.0.1 rc3 release candidate

[Guillaume Rousse]

Also sprach Ian Kent:
> - fix get_query_dn not looking in subtree for LDAP search (missed second
>   occurance).
Well, this patch was mine :)

Actually, I don't really care about getting mentionned in the changelog,
but I'd like my other patches from
http://svn.mandriva.com/cgi-bin/viewvc.cgi/packages/cooker/autofs/releases/5.0.1/0.rc3.1mdv2007.1/SOURCES/
to get in also, especially when they fix objectives troubles:
- autofs-5.0.1-rc1-fix-hesiod-check.patch is needed to build without
hesiod but with other libraries (such as LDAP)
- autofs-5.0.1-rc1-fix-man-page.patch fix an error in a man page, and
avoid a useless substitution

The following are more discussable, and only concern enhancing
configuration readability:
- autofs-5.0.1-rc1-drop-default-prefix-from-config.patch drop useless
DEFAULT_ prefix from configuration directives
- autofs-5.0.1-rc3-separate-config-files.patch split configuration in
two parts, one for the automount daemon, one for the automount init script
- autofs-5.0.1-rc1-cleanup-config-files-names.patch just tidify
configuration files names

Thanks.

Re: [ANNOUNCE] autofs 5.0.1 rc3 release candidate

[Ian Kent]

On Wed, 2007-01-17 at 11:59 +0100, Guillaume Rousse wrote:
> Also sprach Ian Kent:
> > - fix get_query_dn not looking in subtree for LDAP search (missed second
> >   occurance).
> Well, this patch was mine :)

Apologies, you are right.
I do try to credit the source of patches and I didn't mean to leave you
out, sorry.

Now that I think I've just about fixed the more important v5
functionality issues, at least almost all the ones I know about, I'll
have a look through these, thanks.

I must say though that I want to make as few changes as I can till after
the first v5 release. Following that I'll collect the outstanding
patches and apply them and we can deal with any issues.

How urgent are these first two for your needs?

> 
> Actually, I don't really care about getting mentionned in the changelog,
> but I'd like my other patches from
> http://svn.mandriva.com/cgi-bin/viewvc.cgi/packages/cooker/autofs/releases/5.0.1/0.rc3.1mdv2007.1/SOURCES/
> to get in also, especially when they fix objectives troubles:
> - autofs-5.0.1-rc1-fix-hesiod-check.patch is needed to build without
> hesiod but with other libraries (such as LDAP)

Tell me a little more about this please.

> - autofs-5.0.1-rc1-fix-man-page.patch fix an error in a man page, and
> avoid a useless substitution

The Makefile isn't updated by this patch.
Does this build and install without error in your build environment?

> 
> The following are more discussable, and only concern enhancing
> configuration readability:
> - autofs-5.0.1-rc1-drop-default-prefix-from-config.patch drop useless
> DEFAULT_ prefix from configuration directives

Mmmm .. remember that.
In principle it looked OK but I couldn't help thinking there was
something missed in the patch, I didn't know what but you've been using
it for a while now so I must be mistaken. The other thing, which is more
my way of thinking about the meaning of the names, is that I felt that
the names are more meaningful as they are. For example the master map
name can be changed by passing a value on the command line, and so that
changes the configured default, and similarly the timeout.

> - autofs-5.0.1-rc3-separate-config-files.patch split configuration in
> two parts, one for the automount daemon, one for the automount init script

Initially I didn't want to split these up but looking at the patch and
thinking about it I'm starting to warm to the idea.

> - autofs-5.0.1-rc1-cleanup-config-files-names.patch just tidify
> configuration files names

Not sure about changing the config file name.
There are two common naming schemes, one is what your proposing and the
other is what's being used. I come from a background where the later is
most common so I tend to prefer using it. But that probably needs some
more thought and a closer look at the patch.

Ian

Re: [ANNOUNCE] autofs 5.0.1 rc3 release candidate

[Jeff Moyer]

==> On Wed, 17 Jan 2007 11:59:49 +0100, Guillaume Rousse <Guillaume.Rousse@inria.fr> said:

Guillaume> - autofs-5.0.1-rc3-separate-config-files.patch split configuration in
Guillaume> two parts, one for the automount daemon, one for the automount init script

What is the purpose of this?

Re: [ANNOUNCE] autofs 5.0.1 rc3 release candidate

[Guillaume Rousse]

Also sprach Jeff Moyer:
> ==> On Wed, 17 Jan 2007 11:59:49 +0100, Guillaume Rousse <Guillaume.Rousse@inria.fr> said:
> 
> Guillaume> - autofs-5.0.1-rc3-separate-config-files.patch split configuration in
> Guillaume> two parts, one for the automount daemon, one for the automount init script
> 
> What is the purpose of this?
Well, to mimic standard setup for all other services.

I also think separating vendor-specific parts, such as init scripts,
from upstream parts, such as main daemon configuration, make life easier
for both parts.

It's easier for packagers to provide vendor-specific customisation
without side effects (I'd prefer avoiding to add a
USE_MDV_SPECIFIC_FEATURE_X in autofs configuration). And it avoid
upstream maintainer to provide vendor-targeted switches, such as
--where-do-you-want-the-automount-configuration-to-be-installed, as this
 standard (%sysconfdir).

Re: [ANNOUNCE] autofs 5.0.1 rc3 release candidate

[Jeff Moyer]

==> On Wed, 17 Jan 2007 15:00:18 +0100, Guillaume Rousse <Guillaume.Rousse@inria.fr> said:

Guillaume> Also sprach Jeff Moyer:
Guillaume> > ==> On Wed, 17 Jan 2007 11:59:49 +0100, Guillaume Rousse <Guillaume.Rousse@inria.fr> said:
Guillaume> > 
Guillaume> > Guillaume> - autofs-5.0.1-rc3-separate-config-files.patch split configuration in
Guillaume> > Guillaume> two parts, one for the automount daemon, one for the automount init script
Guillaume> > 
Guillaume> > What is the purpose of this?
Guillaume> Well, to mimic standard setup for all other services.

Guillaume> I also think separating vendor-specific parts, such as init scripts,
Guillaume> from upstream parts, such as main daemon configuration, make life easier
Guillaume> for both parts.

Guillaume> It's easier for packagers to provide vendor-specific customisation
Guillaume> without side effects (I'd prefer avoiding to add a
Guillaume> USE_MDV_SPECIFIC_FEATURE_X in autofs configuration). And it avoid
Guillaume> upstream maintainer to provide vendor-targeted switches, such as
Guillaume> --where-do-you-want-the-automount-configuration-to-be-installed, as this
Guillaume>  standard (%sysconfdir).

OK, I'm with you.  Ian, I'm in favor of this change.

-Jeff

Re: [ANNOUNCE] autofs 5.0.1 rc3 release candidate

[Ian Kent]

On Wed, 2007-01-17 at 09:08 -0500, Jeff Moyer wrote:
> ==> On Wed, 17 Jan 2007 15:00:18 +0100, Guillaume Rousse <Guillaume.Rousse@inria.fr> said:
> 
> Guillaume> Also sprach Jeff Moyer:
> Guillaume> > ==> On Wed, 17 Jan 2007 11:59:49 +0100, Guillaume Rousse <Guillaume.Rousse@inria.fr> said:
> Guillaume> > 
> Guillaume> > Guillaume> - autofs-5.0.1-rc3-separate-config-files.patch split configuration in
> Guillaume> > Guillaume> two parts, one for the automount daemon, one for the automount init script
> Guillaume> > 
> Guillaume> > What is the purpose of this?
> Guillaume> Well, to mimic standard setup for all other services.
> 
> Guillaume> I also think separating vendor-specific parts, such as init scripts,
> Guillaume> from upstream parts, such as main daemon configuration, make life easier
> Guillaume> for both parts.
> 
> Guillaume> It's easier for packagers to provide vendor-specific customisation
> Guillaume> without side effects (I'd prefer avoiding to add a
> Guillaume> USE_MDV_SPECIFIC_FEATURE_X in autofs configuration). And it avoid
> Guillaume> upstream maintainer to provide vendor-targeted switches, such as
> Guillaume> --where-do-you-want-the-automount-configuration-to-be-installed, as this
> Guillaume>  standard (%sysconfdir).
> 
> OK, I'm with you.  Ian, I'm in favor of this change.

Yep.
As I say the suggestion is growing on me.
I want to look a bit closer at the patch as well.

Ian